摘要
目前,船舶轨迹预测存在数据噪声严重、缺乏考虑历史轨迹的相似性的问题,导致预测精度不高,难以满足实际需求。针对该问题,从船舶自动识别系统(Automatic Identification System,AIS)数据去噪、在预测模型中考虑历史数据两个方面提高船舶轨迹单步预测的精度和可靠性。根据相邻时刻的AIS数据修复当前时刻的船舶运动参数;使用二维长短时记忆网络(Long Short-Term Memory,LSTM)算法分别从时间和空间的角度将当前船舶的航迹数据和历史轨迹数据相融合,建立航迹预测模型和区域预测模型;利用试验数据验证模型的综合性能。试验结果表明:该模型可获得精度较高的船舶航行轨迹。
The accuracy of one-step-ahead prediction is improved in two ways:reducing noise interference by the aid of AIS(Automatic Identification System)data and introducing historical information to the prediction process.The current motion measurements are corrected according to AIS information at adjacent time and then fused with historical track in terms of space and time with LSTM(Long Short-Term Memory)algorithm to get final prediction.The track prediction model and area prediction model are built.The models are verified through simulation.
作者
高大为
朱永生
张金奋
鄢博冉
何延康
闫柯
GAO Dawei;ZHU Yongsheng;ZHANG Jinfen;YAN Boran;HE Yankang;YAN Ke(Key Laboratory of Education Ministry for Modern Design and Rotor-Bearing System,Xi’an Jiaotong University,Xi’an 710049,China;Intelligent Transportation Systems Research Center,Wuhan University of Technology,Wuhan 430063,China)
出处
《中国航海》
CSCD
北大核心
2021年第3期56-63,共8页
Navigation of China
基金
国家重点研发计划项目(2017YFC0804900,2017YFC0804904)
